Abstract:
In view of condition of complex dynamic scene changes in coal mine intelligent video monitoring environment, three important steps in moving target detection were researched which were background modeling and updating, foreground detecting, motion shadow detecting and removing. For the three steps, corresponding processing methods were put forward: a self-adaptive background modeling and updating method based on IFCM clustering algorithm was proposed, the method was used to unsupervised clustering of pixels' gray values, different number of clusters was adaptively selected to construct the pixels' background model, and automatic updating of the background model was completed by modifying, adding and deleting clusters with the scenes' change; a foreground detection method was proposed which combined the background difference, three frame difference and spatial neighborhood information, so accurate foreground targets were obtained; a motion shadow detection and removal method was proposed on the basis that the pixels' of gray images have characteristics of luminance correlation and texture invariance before and after shadow covering, so detecting and removing moving shadow was realized. The experimental results verify effectiveness and superiority of the method.
